The manufacturing plant is viewed as a group of interacting subsystems. The research described in this paper is aimed at characterizing the subsystem interactions such that an intelligent system can make appropriate decisions in the face of shop floor contingencies. The goal of such decision making is to insure the meeting of global production goals by making real-time tradeoffs at the level of local subsystem goals. Manufacturing subsystems include receiving, shipping, production lines, maintenance, and material handling. It is assumed that subsystem interactions can be captured as schedule interactions. Active rescheduling is the use of manufacturing parameters as "cues" for triggering subsystem rescheduling. A distributed agent system is described, in which an agent represents a subsystem scheduler. Two agendas are employed: one for the posting of proposed schedule revisions and a second agenda for the processing of contingencies which must be resolved before proposed schedule revisions can be accepted.